From 9b7eecdcfeb943f130d86bbc249fde4994b6fe30 Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Thu, 22 Sep 2011 21:50:15 -0400 Subject: pnfs: recoalesce when ld read pagelist fails For pnfs pagelist read failure, we need to pg_recoalesce and resend IO to mds. Signed-off-by: Peng Tao Signed-off-by: Jim Rees Cc: stable@kernel.org [3.0] Signed-off-by: Trond Myklebust --- fs/nfs/read.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'fs/nfs/read.c') diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 2171c043ab08..bfc20b160243 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -541,13 +541,23 @@ static void nfs_readpage_result_full(struct rpc_task *task, void *calldata) static void nfs_readpage_release_full(void *calldata) { struct nfs_read_data *data = calldata; + struct nfs_pageio_descriptor pgio; + if (data->pnfs_error) { + nfs_pageio_init_read_mds(&pgio, data->inode); + pgio.pg_recoalesce = 1; + } while (!list_empty(&data->pages)) { struct nfs_page *req = nfs_list_entry(data->pages.next); nfs_list_remove_request(req); - nfs_readpage_release(req); + if (!data->pnfs_error) + nfs_readpage_release(req); + else + nfs_pageio_add_request(&pgio, req); } + if (data->pnfs_error) + nfs_pageio_complete(&pgio); nfs_readdata_release(calldata); } -- cgit v1.2.3 From d00c5d43866720963a265fa3129f3203cac35b8e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Oct 2011 12:17:29 -0700 Subject: NFS: Get rid of nfs_restart_rpc() It can trivially be replaced with rpc_restart_call_prepare. Signed-off-by: Trond Myklebust --- fs/nfs/internal.h | 10 ---------- fs/nfs/nfs4filelayout.c | 14 ++++---------- fs/nfs/nfs4proc.c | 17 ++++++++--------- fs/nfs/read.c | 2 +- fs/nfs/unlink.c | 4 ++-- fs/nfs/write.c | 2 +- net/sunrpc/clnt.c | 4 +++- 7 files changed, 19 insertions(+), 34 deletions(-) (limited to 'fs/nfs/read.c') diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index ab12913dd473..c1a1bd8ddf1c 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -457,13 +457,3 @@ unsigned int nfs_page_array_len(unsigned int base, size_t len) PAGE_SIZE - 1) >> PAGE_SHIFT; } -/* - * Helper for restarting RPC calls in the possible presence of NFSv4.1 - * sessions. - */ -static inline int nfs_restart_rpc(struct rpc_task *task, const struct nfs_client *clp) -{ - if (nfs4_has_session(clp)) - return rpc_restart_call_prepare(task); - return rpc_restart_call(task); -} diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 4c78c62639e6..09119418402f 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -122,7 +122,6 @@ static int filelayout_async_handle_error(struct rpc_task *task, static int filelayout_read_done_cb(struct rpc_task *task, struct nfs_read_data *data) { - struct nfs_client *clp = data->ds_clp; int reset = 0; dprintk("%s DS read\n", __func__); @@ -134,9 +133,8 @@ static int filelayout_read_done_cb(struct rpc_task *task, if (reset) { pnfs_set_lo_fail(data->lseg); nfs4_reset_read(task, data); - clp = NFS_SERVER(data->inode)->nfs_client; } - nfs_restart_rpc(task, clp); + rpc_restart_call_prepare(task); return -EAGAIN; } @@ -203,17 +201,13 @@ static int filelayout_write_done_cb(struct rpc_task *task, if (filelayout_async_handle_error(task, data->args.context->state, data->ds_clp, &reset) == -EAGAIN) { - struct nfs_client *clp; - dprintk("%s calling restart ds_clp %p ds_clp->cl_session %p\n", __func__, data->ds_clp, data->ds_clp->cl_session); if (reset) { pnfs_set_lo_fail(data->lseg); nfs4_reset_write(task, data); - clp = NFS_SERVER(data->inode)->nfs_client; - } else - clp = data->ds_clp; - nfs_restart_rpc(task, clp); + } + rpc_restart_call_prepare(task); return -EAGAIN; } @@ -245,7 +239,7 @@ static int filelayout_commit_done_cb(struct rpc_task *task, prepare_to_resend_writes(data); pnfs_set_lo_fail(data->lseg); } else - nfs_restart_rpc(task, data->ds_clp); + rpc_restart_call_prepare(task); return -EAGAIN; } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index ba0da50865fe..d2ae413c986a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3191,7 +3191,7 @@ static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data) struct nfs_server *server = NFS_SERVER(data->inode); if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) { - nfs_restart_rpc(task, server->nfs_client); + rpc_restart_call_prepare(task); return -EAGAIN; } @@ -3241,7 +3241,7 @@ static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data struct inode *inode = data->inode; if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) { - nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client); + rpc_restart_call_prepare(task); return -EAGAIN; } if (task->tk_status >= 0) { @@ -3298,7 +3298,7 @@ static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_write_data *dat struct inode *inode = data->inode; if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) { - nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client); + rpc_restart_call_prepare(task); return -EAGAIN; } nfs_refresh_inode(inode, data->res.fattr); @@ -3838,7 +3838,7 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata) default: if (nfs4_async_handle_error(task, data->res.server, NULL) == -EAGAIN) { - nfs_restart_rpc(task, data->res.server->nfs_client); + rpc_restart_call_prepare(task); return; } } @@ -4092,8 +4092,7 @@ static void nfs4_locku_done(struct rpc_task *task, void *data) break; default: if (nfs4_async_handle_error(task, calldata->server, NULL) == -EAGAIN) - nfs_restart_rpc(task, - calldata->server->nfs_client); + rpc_restart_call_prepare(task); } } @@ -4926,7 +4925,7 @@ static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata) task->tk_status = 0; /* fall through */ case -NFS4ERR_RETRY_UNCACHED_REP: - nfs_restart_rpc(task, data->clp); + rpc_restart_call_prepare(task); return; } dprintk("<-- %s\n", __func__); @@ -5767,7 +5766,7 @@ static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata) server = NFS_SERVER(lrp->args.inode); if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) { - nfs_restart_rpc(task, lrp->clp); + rpc_restart_call_prepare(task); return; } spin_lock(&lo->plh_inode->i_lock); @@ -5938,7 +5937,7 @@ nfs4_layoutcommit_done(struct rpc_task *task, void *calldata) } if (nfs4_async_handle_error(task, server, NULL) == -EAGAIN) { - nfs_restart_rpc(task, server->nfs_client); + rpc_restart_call_prepare(task); return; } diff --git a/fs/nfs/read.c b/fs/nfs/read.c index bfc20b160243..e866a7e6e2d5 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -435,7 +435,7 @@ static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data argp->offset += resp->count; argp->pgbase += resp->count; argp->count -= resp->count; - nfs_restart_rpc(task, NFS_SERVER(data->inode)->nfs_client); + rpc_restart_call_prepare(task); } /* diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c index b2fbbde58e44..4f9319a2e567 100644 --- a/fs/nfs/unlink.c +++ b/fs/nfs/unlink.c @@ -87,7 +87,7 @@ static void nfs_async_unlink_done(struct rpc_task *task, void *calldata) struct inode *dir = data->dir; if (!NFS_PROTO(dir)->unlink_done(task, dir)) - nfs_restart_rpc(task, NFS_SERVER(dir)->nfs_client); + rpc_restart_call_prepare(task); } /** @@ -369,7 +369,7 @@ static void nfs_async_rename_done(struct rpc_task *task, void *calldata) struct dentry *new_dentry = data->new_dentry; if (!NFS_PROTO(old_dir)->rename_done(task, old_dir, new_dir)) { - nfs_restart_rpc(task, NFS_SERVER(old_dir)->nfs_client); + rpc_restart_call_prepare(task); return; } diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 2084a6494218..ad90b0c998cb 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1305,7 +1305,7 @@ void nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data) */ argp->stable = NFS_FILE_SYNC; } - nfs_restart_rpc(task, server->nfs_client); + rpc_restart_call_prepare(task); return; } if (time_before(complain, jiffies)) { diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index c5347d29cfb7..f0268ea7e711 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -850,7 +850,9 @@ rpc_restart_call_prepare(struct rpc_task *task) { if (RPC_ASSASSINATED(task)) return 0; - task->tk_action = rpc_prepare_task; + task->tk_action = call_start; + if (task->tk_ops->rpc_call_prepare != NULL) + task->tk_action = rpc_prepare_task; return 1; } EXPORT_SYMBOL_GPL(rpc_restart_call_prepare); -- cgit v1.2.3 From fbb5a9abf0d589e9471dc93b18025b7b921d22c9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Oct 2011 12:17:29 -0700 Subject: NFS: Get rid of unnecessary calls to ClearPageError() in read code The generic file read code does that for us anyway. Signed-off-by: Trond Myklebust --- fs/nfs/read.c | 2 -- fs/nfs/write.c | 1 - 2 files changed, 3 deletions(-) (limited to 'fs/nfs/read.c') diff --git a/fs/nfs/read.c b/fs/nfs/read.c index e866a7e6e2d5..09829d96d207 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -322,7 +322,6 @@ static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head offset += len; } while(nbytes != 0); atomic_set(&req->wb_complete, requests); - ClearPageError(page); desc->pg_rpc_callops = &nfs_read_partial_ops; return ret; out_bad: @@ -357,7 +356,6 @@ static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, struct list_head * req = nfs_list_entry(head->next); nfs_list_remove_request(req); nfs_list_add_request(req, &data->pages); - ClearPageError(req->wb_page); *pages++ = req->wb_page; } req = nfs_list_entry(data->pages.next); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index ad90b0c998cb..2219c88d96b2 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1011,7 +1011,6 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc, struct list_head *r req = nfs_list_entry(head->next); nfs_list_remove_request(req); nfs_list_add_request(req, &data->pages); - ClearPageError(req->wb_page); *pages++ = req->wb_page; } req = nfs_list_entry(data->pages.next); -- cgit v1.2.3 From fba730050d1246d0e6ef44e026e0b584732fec2b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Oct 2011 12:17:29 -0700 Subject: NFS: Don't rely on PageError in nfs_readpage_release_partial Don't rely on the PageError flag to tell us if one of the partial reads of the page failed. Instead, replace that with a dedicated flag in the struct nfs_page. Then clean out redundant uses of the PageError flag: the VM no longer checks it for reads. Signed-off-by: Trond Myklebust --- fs/nfs/read.c | 7 ++----- include/linux/nfs_page.h | 1 + 2 files changed, 3 insertions(+), 5 deletions(-) (limited to 'fs/nfs/read.c') diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 09829d96d207..fd58e909842b 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -276,7 +276,6 @@ nfs_async_read_error(struct list_head *head) while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - SetPageError(req->wb_page); nfs_readpage_release(req); } } @@ -330,7 +329,6 @@ out_bad: list_del(&data->list); nfs_readdata_free(data); } - SetPageError(page); nfs_readpage_release(req); return -ENOMEM; } @@ -460,10 +458,10 @@ static void nfs_readpage_release_partial(void *calldata) int status = data->task.tk_status; if (status < 0) - SetPageError(page); + set_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags); if (atomic_dec_and_test(&req->wb_complete)) { - if (!PageError(page)) + if (!test_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags)) SetPageUptodate(page); nfs_readpage_release(req); } @@ -656,7 +654,6 @@ readpage_async_filler(void *data, struct page *page) return 0; out_error: error = PTR_ERR(new); - SetPageError(page); out_unlock: unlock_page(page); return error; diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index e2791a27a901..ab465fe8c3d6 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -34,6 +34,7 @@ enum { PG_NEED_COMMIT, PG_NEED_RESCHED, PG_PNFS_COMMIT, + PG_PARTIAL_READ_FAILED, }; struct nfs_inode; -- cgit v1.2.3 From b6ee8cd2642f6d822dd1a4ba62298b65ff99b72e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Oct 2011 12:17:29 -0700 Subject: NFS: Get rid of the nfs_rdata_mempool We don't need a mempool in order to guarantee reliable NFS read performance. Signed-off-by: Trond Myklebust --- fs/nfs/read.c | 17 ++++------------- 1 file changed, 4 insertions(+), 13 deletions(-) (limited to 'fs/nfs/read.c') diff --git a/fs/nfs/read.c b/fs/nfs/read.c index fd58e909842b..8b48ec63f722 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -35,16 +35,13 @@ static const struct rpc_call_ops nfs_read_partial_ops; static const struct rpc_call_ops nfs_read_full_ops; static struct kmem_cache *nfs_rdata_cachep; -static mempool_t *nfs_rdata_mempool; - -#define MIN_POOL_READ (32) struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount) { - struct nfs_read_data *p = mempool_alloc(nfs_rdata_mempool, GFP_KERNEL); + struct nfs_read_data *p; + p = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); if (p) { - memset(p, 0, sizeof(*p)); INIT_LIST_HEAD(&p->pages); p->npages = pagecount; if (pagecount <= ARRAY_SIZE(p->page_array)) @@ -52,7 +49,7 @@ struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount) else { p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); if (!p->pagevec) { - mempool_free(p, nfs_rdata_mempool); + kmem_cache_free(nfs_rdata_cachep, p); p = NULL; } } @@ -64,7 +61,7 @@ void nfs_readdata_free(struct nfs_read_data *p) { if (p && (p->pagevec != &p->page_array[0])) kfree(p->pagevec); - mempool_free(p, nfs_rdata_mempool); + kmem_cache_free(nfs_rdata_cachep, p); } void nfs_readdata_release(struct nfs_read_data *rdata) @@ -716,16 +713,10 @@ int __init nfs_init_readpagecache(void) if (nfs_rdata_cachep == NULL) return -ENOMEM; - nfs_rdata_mempool = mempool_create_slab_pool(MIN_POOL_READ, - nfs_rdata_cachep); - if (nfs_rdata_mempool == NULL) - return -ENOMEM; - return 0; } void nfs_destroy_readpagecache(void) { - mempool_destroy(nfs_rdata_mempool); kmem_cache_destroy(nfs_rdata_cachep); } -- cgit v1.2.3